Apache Kafka Consumer Backends

You can configure the AppDynamics Java Agent to detect entry points for Apache Kafka consumer activity using KafkaConsumer.poll() (introduced in Kafka v0.10) and Kafka SimpleConsumer.fetch() (introduced in Kafka v0.9). Prior to version 0.11, the Kafka payload did not include a location to store correlation data, so end-to-end Business Transaction correlation is only possible with the Kafka client and broker >= 0.11.

KafkaConsumer.poll

To instrument Kafka consumer entry points using KafkaConsumer.poll(), identify the method in which the consumer reads messages in a loop with a custom interceptor definition. We instrument the iterator's next method to start and end the Business Transaction for each message. There could be many iterators used for iterating messages but we only support iterators of the following types:

• kafka.consumer.ConsumerIterator
• org.apache.kafka.clients.consumer.ConsumerRecords$ConcatenatedIterable$1

1. Identify the class and method of the loop that processes messages from Kafka. Consider, for example, a class MyConsumer that employs the following loop to poll and process messages from Kafka:

   ConsumerRecords<String, String> records = kafkaConsumer.poll(1000);
   pollMessages(records);
   private void pollMessages(ConsumerRecords<String, String>records) throws Exception {
      //AppDynamics instrumentation gets applied here
   for (ConsumerRecord<String, String> record : records) {
      //Processing of the records
      System.out.println(record.value());
      }
   }

   CODE

   For this case, you want to intercept:

   • Class: MyConsumer
   • Method: pollMessages

   The interceptor can also be applied to a method that processes individual records, not just a loop. For example:

   ConsumerRecords<String, String> records = kafkaConsumer.poll(1000);
   pollMessages(records);
   private void pollMessages(ConsumerRecords<String, String>records) throws Exception {
      //AppDynamics instrumentation gets applied here
   for (ConsumerRecord<String, String> record : records) {
      processRecord(record)
      }
   }

   CODE
2. Use your preferred text editor to create and edit a file named custom-interceptors.xml at the following path:
   <agent_home>/<version_number>/conf 
   
   For example:
   /usr/home/appdynamics/appagent/ver4.3.1.0/conf/custom-interceptors.xml

3. Copy the following XML to custom-interceptors.xml:

   <custom-interceptors>
       <custom-interceptor>
           <interceptor-class-name>com.singularity.KafkaMarkerMethodInterceptor</interceptor-class-name>
           <match-class type="matches-class">
               <name filter-type="equals">my-fully-qualified-class-name</name>
           </match-class>
           <match-method>
               <name>my-method-name</name>
           </match-method>
       </custom-interceptor>
   </custom-interceptors>

   XML

4. Set the value of the class name to the name of your consumer class. For instance, to specify the MyConsumer class:

   <match-class type="matches-class">
       <name filter-type="equals">com.mycompany.mypackage.MyConsumer</name>
   </match-class>

   XML

5. Set the value of the method name to the name of your message processing loop method. For instance, to specify the pollMessages method:

   <match-method>
       <name>pollMessages</name>
   </match-method>

   XML

   After the Java Agent reads the updated configuration, it detects consumer activity and upstream Kafka queue. The application flow map shows the tier receiving data from the Kafka queue.

Kafka SimpleConsumer Entry Points

To enable consumer entry points for Kafka clients that retrieve messages using SimpleConsumer.fetch(), register the enable-kafka-consumer node property with a value of "true."

Kafka consumer activity shows up as an exit call in this case.